Usefulness of Power Consumption Simulated Data of Inhabited Houses for Abnormal Activity Detection

Smart devices and sensors can become a powerful ally for detecting strange behaviour patterns from elderly people in Activities of Daily Living (ADLs), which are often associated with cognitive impairment or dementia. Our proposal focuses on devices able to measure power consumption, because they are influenced by a wide variety of Instrumental ADLs, they are not expensive and they are easy to deploy. For studying how to find useful patterns in energy consumption data, first, it needs to be gathered in controlled environments. However obtaining data from elderly people homes correctly labelled can be challenging and present ethical issues. As a solution, this paper proposes simulating an inhabited sensorized home as a digital shadow using the open source tool SiSHoDiT, which has been improved for this article to be able to generate data from sensors and/or devices which a fixed sample rate. The synthetic data usefulness is evaluated by performing an analysis similar that could also be applied to real data. The findings demonstrate that simulation is a viable solution for generating realistic, context-rich datasets that can help in the understanding of IADLs and related abnormal behaviour detection, although synthetic data is cleaner and easiest to analyse. This research contributes to the fields of ambient assisted living, smart home technology, and appliance load monitoring; and opens new avenues for future innovations in digital twin technologies and their applications in real-time monitoring, anomaly detection, and improvement of life quality of people at home.